next article | contents

IMAGE PROBLEM

---

A Sussex professor is part of a group that has secured a grant to study the interpretation of magnetic resonance brain images.

Prof Benedict du Boulay (COGS) described the two main objectives of the research:- to investigate the nature of diagnosis in this complex domain, and to further develop an existing prototype decision support and training aid, based on previous work which has been funded by an ESRC cognitive engineering grant.

Radiological diagnosis and decision- making is a particularly difficult task. It involves the extraction and interpretation of information from complex medical images and reasoning about competing pathologies that often exhibit similar abnormal features in the images.

A senior staff radiologist sees in the order of 80 to 100 cases per week, or around 100,000 in a working life. The team will focus on helping radiologists to learn to assess the areas of uncertainty and imprecision using visualisation techniques. They will also improve the precision of an 'image description language' which is used to describe case images in a database, and then display their distribution in an 'overview space' in the decision-support and training tool.

The difficulty of interpreting these images is compounded by the following factors:

• the high level of detail available in magnetic resonance images,
• the sensitivity of the appearance of the image to the initial settings of the scanner,
• the great variability of presentation, even among normal patients,
• the partial understanding, even among experts, of the relationship between the appearance of brain lesions and the nature of the underlying pathology.

Benedict will be working with Prof Mike Sharples and Dr Nathan Jeffery, both ex-COGS and now at the University of Birmingham, Prof Derek Teather and Dr Briony Teather of De Montfort University, and Prof George du Boulay of the Institute of Neurology.

next article | contents

Friday February 27th 1998